A second instrument

Does reported experience have an order?

The field offers static correlates of whether an experience is present. No entry in the catalog carries a tested account of how one unfolds. Structural Metaphysical Dynamics codes four content-defined elements in first-person near-death narratives, between the body at the start and the physical return at the end. It asks one question: do people who have nothing in common report them in the same order?

Seeing→Passage→Adventure→Boundary

The four elements are content-defined: each is individuated by what is reported, not by when it is reported. That distinction is load-bearing. A sequence test cannot recover order from states whose coding rule already references order, so the frame of the narrative — the body at the start, the physical return at the end — is held outside the test and carries no evidential weight for the order claim.

Oneness is extracted by the protocol but excluded from the primary test.

What was found

Discovery

76.7%

mean account-level concordance

653 accounts, against a 50% null. Not one of 1,000,000 order-shuffled corpora reached it.

Held out

80.0%

concordance, order frozen

The order was then frozen and applied to a separate raw corpus: 434 narratives, 165 of them eligible. Cohen’s d of about 0.84.

Independent raters

94.9%

forward share, blind reviewers

Five reviewers, blind to the predicted order and to every machine label, coded the same 47 accounts: 168 forward transitions against 9, against a null of 0.50. The minimum results were specified in advance and were met.

The three panels are one measurement read left to right, not three different claims. The same concordance statistic is computed the same way in each: on the corpus the order was found in, then on a separate corpus with the order frozen, then by human coders reading the raw narratives with the model's coding withheld.

SMD measures the order of reports. It makes no claim that anything seen was seen accurately, and the theory on this board cites no veridical case as support.

The thresholds the panel was held to

They were fixed before the data existed, and the panel was scored against them rather than against whatever it happened to return:

Provisional≥ 75% agreement, or κ ≥ 0.60

Strong≥ 80–85% agreement, or κ ≥ 0.70

Matrixhuman-coded transitions within ±0.10 of the primary matrix, cell by cell

The minimum results specified in advance were met. They were not adjusted afterwards.

This is not a measure of how closely human coders agree with a model, and it is not a measure or certification of model accuracy. It is the same protocol hypothesis, tested by a different kind of operator — which is the only design that separates structure in the text from a template in the model.

What has not been done

Independent laboratory replication. Not attempted. Invited.

Comparison against other altered states. Existing comparisons between near-death and pharmacologically induced experience have measured content rather than order. No comparable sequential findings have been reported from non-transcendental altered states. This is identified as a further decisive open test, and it is set out in full under limitations.

Stricter nulls. The executed null shuffles the order of reported elements within each account. A permutation ensemble cannot resolve a probability below roughly one over the number of shuffles, so the result is reported as a bound — no one of 1,000,000 shuffled corpora reached the observed concordance — rather than as a point estimate beyond that resolution.